The Cybernetic Brain

by Andrew Pickering

  The History of Musicolour

  In the early 1950s, first at Cambridge and then at System Research, Gordon and Elizabeth Pask and Robin McKinnon-Wood sought to turn Musicolour into a commercial proposition, beginning in 1953 at the Pomegranate Club in Cambridge—"an eclectically Dadaist organisation"—followed by "a bizarre and eventful tour of the north country" and an eventual move to London (Pask 1971, 78). McKinnon-Wood (1993, 131) later reminisced, "I think my favourite memory of MusiColour was the time we demonstrated the portable version to Mr Billy Butlin [the proprietor of Butlin's holiday camps] in his office. . . . Shortly after his arrival it exploded in a cloud of white smoke. . . . I switched it back on again and it worked perfectly." The first London performance was at Bolton's Theatre and took a traditionally theatrical form. In a piece called "Moon Music," a musician played, Musicolour modulated the lights on a stage set, and, to liven things up, marionettes danced on stage. The marionettes were supposed to move in synchrony with the lights but instead dismembered themselves over the audience (Pask 1971, 82–83). The next show was at Valerie Hovenden's Theatre Club in the crypt of St. Anne's Church on Dean Street. There, in a piece called "Nocturne," attempts were made to link the motions of a human dancer into Musicolour's input—"this proved technically difficult but the aesthetic possibilities are indisputable" (86). Then (86), "since the system was costly to maintain and since the returns were modest, the Musicolour enterprise fell into debt. We secured inexpensive premises above the King's Arms in Tabernacle Street which is a curiously dingy part of the City of London, often engulfed in a sort of beer-sodden mist. There, we set up the system and tried to sell it in any possible way: at one extreme as a pure art form, at the other as an attachment for juke boxes." The story then passed through Churchill's Club, where waiters "dropped cutlery into its entrails [but] the audience reaction was favorable and Musicolour became a permanent feature of the spectacle." After that, Musicolour was used to drive the 120 kilowatt lights at the Mecca Locarno dance hall in Streatham, where, alas, "it became clear that in large scale (and commercially viable) situations, it was difficult or impossible to make genuine use of the system."10 "Musicolour made its last appearance in 1957, at a ball organized by Michael Gillis. We used a big machine, a small machine and a collection of display media accumulated over the years. But there were other things to do. After the ball, in the crisp, but fragrant air of St. James's Park, the Musicolour idea was formally shelved. I still have a small machine. But it does not work any longer and is of chiefly sentimental value" (Pask 1971, 86–88). We can follow the subsequent mutations of Musicolour in Pask's career below, but one other aspect of its historical development is worth mentioning. As mentioned above by Elizabeth Pask, Gordon had an enduring interest in learning, and we should see how Musicolour fitted into this. The point to note is that in performance the performer learned (performatively rather than cognitively) about the machine (and vice versa), and Pask therefore regarded Musicolour as a machine in which one could learn—scientifically, in a conventional sense—about learning. Thus, in the show at Bolton's Theatre (Pask 1971, 83, 85–86),

  it was possible to investigate the stability of the coupling [between performer and machine]. In this study arbitrary disturbances were introduced into the feedback loop wihout the performer's knowledge. Even though he is ignorant of their occurrence, these disturbances are peculiarly distracting to the performer, who eventually becomes infuriated and opts out of the situation. But there is an inherent stability in the man-machine relation which allows the performer to tolerate a certain level of disturbance. We found that the tolerable level increases as the rapport is established (up to a limit of one hour at any rate). . . . Meanwhile, John Clark, a psychiatrist, had come to the theatre and we jointly observed some phenomena related to the establishment of rapport. First, there is a loss of time sense on the performer's part. One performer, for example, tootled away on his instrument from 10 p.m. to 5 a.m. and seemed unaware that much time had passed; an hour, he thought, at the most. This effect . . . was ubiquitous. Next, there is a group of phenomena bearing on the way in which performers train the machine. As a rule, the performer starts off with simple tricks which are entirely open to description. He says, for example, that he is accenting a chord in a particular passage in order to associate a figure in the display with high notes. . . . Soon . . . the determinate trick gives way to a behaviour pattern which the performer cannot describe but which he adopts to achieve a well-defined goal. Later still, the man-machine interaction takes place at a higher level of abstraction. Goals are no longer tied to properties as sensed by the property filters (though, presumably, they are tied to patterns of properties). From the performer's point of view, training becomes a matter of persuading the machine to adopt a visual style which fits the mood of his performance. At this stage . . . the performer conceives the machine as an extension of himself, rather than as a detached or disassociated entity.

  In this sense, Musicolour was, for Pask, an early venture into the experimental psychology of learning and adaptation which led eventually to his 1964 PhD in psychology. I am not going to try to follow this scientific work here, since there was nothing especially cybernetic about it, but we should bear it in mind in the later discussion of Pask's work on training and teaching machines.

  Musicolour and Ontology

  Musicolour was literally a theatrical object; we can also read it as another piece of ontological theater, in the usual double sense. It staged and dramatized the generic form of the cybernetic ontology; at the same time, it exemplified how one might go on, now in the world of theater and aesthetics, if one subscribed to that ontology. Thus, a Musicolour performance staged the encounter of two exceedingly complex systems—the human performer and the machine (we can come back to the latter)—each having its own endogenous dynamics but nevertheless capable of consequential performative interaction with the other in a dance of agency. The human performance certainly affected the output of the machine, but not in a linear and predictable fashion, so the output of the machine fed back to influence the continuing human performance, and so on around the loop and through the duration of the performance. We are reminded here, as in the case of Beer's cybernetics, of the symmetric version of Ashby's multihomeostat setups, and, like Beer's work and Bateson and Laing's, Pask's cybernetic career was characterized by this symmetric vision.

  Beyond this basic observation, we can note that as ontological theater a Musicolour performance undercut any familiar dualist distinction between the human and the nonhuman. The human did not control the performance, nor did the machine. As Pask put it, the performer "trained the machine and it played a game with him. In this sense, the system acted as an extension of the performer with which he could cooperate to achieve effects that he could not achieve on his own" (1971, 78). A Musicolour performance was thus a joint product of a human-machine assemblage. Ontologically, the invitation, as usual, is to think of the world like that—at least the segments that concern us humans, and by analogical extension to the multiplicity of nonhuman elements. This again takes us back to questions of power, which will surface throughout this chapter. In contrast to the traditional impulse to dominate aesthetic media, the Musicolour machine thematized cooperation and revealing in Heidegger's sense. Just as we found Brian Eno "riding the algorithms" in his music in the previous chapter, a Musicolour performer rode the inscrutable dynamics of the machine's circuitry. That is why I said Eno should have read Pask at the end of the previous chapter.

  Second, we can note that as ontological theater Musicolour went beyond some of the limitations of the homeostat. If the homeostat only had twenty- five pregiven states of its uniselector, Musicolour's human component had available an endlessly open-ended range of possibilities to explore, and, inasmuch as the machine adapted and reacted to these, so did the machine. (Of course, unlike Ashby, Pask was not trying to build a freestanding electromechanical brain—his task was much easier in this respect: he could rely on the human performe
r to inject the requisite variety.) At the same time, unlike the homeostat, a Musicolour performance had no fixed goal beyond the very general one of achieving some synesthetic effect, and Pask made no claim to understanding what was required for this. Instead (Pask and McKinnon-Wood 1965, 952),

  other modalities (the best known, perhaps, is Disney's film "Fantasia") have entailed the assumption of a predetermined "synaesthetic" relation. The novelty and scientific interest of this system [Musicolour] emerges from the fact that this assumption is not made. On the contrary, we suppose that the relation which undoubtedly exists between sound (or sound pattern) and light (or light pattern) is entirely personal and that, for a given individual, it is learned throughout a performance. Hence the machine which translates between sound and vision must be a malleable or "learning" device that the performer can "train" (by varying his performance) until it assumes the characteristics of his personally ideal translator.

  The Musicolour performer had to find out what constituted a synesthetic relation between sound and light and how to achieve it. We could speak here of a search process and the temporal emergence of desire—another Heideggerian revealing—rather than of a preconceived goal that governs a performance. In both of these senses, Musicolour constituted a much richer and more suggestive act of ontological theater than the homeostat, though remaining clearly in the homeostatic lineage.

  One subtlety remains to be discussed. I just described Musicolour as one of Beer's exceedingly complex systems. This seems evidently right to me, at least from the perspective of the performer. Even from the simplified description I have given of its functioning, it seems clear that one could not think one's way through Musicolour, anticipating its every adaptation to an evolving sequence of inputs, and this becomes even clearer if one reads Pask's description of all the subtleties in wiring and logic (1971, 78–80). But still, there was a wiring diagram for Musicolour which anyone with a bit of training in electrical engineering could read. So we have, as it were, two descriptions of Musicolour: as an exceedingly complex system (as experienced in practice) and as actually quite simple and comprehensible (as described by its wiring diagram). I am reminded of Arthur Stanley Eddington's two tables: the solid wooden one at which he wrote, and the table as described by physics, made of electrons and nuclei, but mainly empty space. What should we make of this? First, we could think back to the discussion of earlier chapters. In chapter 4 I discussed cellular automata as ontological icons, as exemplifications of the fact that even very simple systems can display enormously complex behavior—as the kinds of objects that might help one imagine the cybernetic ontology more generally. One can think of Musicolour similarly, as a material counterpart to those mathematical systems—thus setting it in the lineage running from the tortoise to DAMS. And second, we could take Musicolour as a reminder that representational understandings of inner workings can often be of little use in our interactions with the world. Though the workings of Musicolour were transparent in the wiring diagram, the best way to get on with it was just to play it. The detour through representation does not rescue us here from the domain of performance.11

  Ontology and Aesthetics

  As one might expect from a cybernetician with roots in the theater, ontology and aesthetics intertwined in Pask's work. I have been quoting at length from an essay Pask wrote in 1968 on Musicolour and its successor, the Colloquy of Mobiles (Pask 1971), which begins with the remark that (76) "man is prone to seek novelty in his environment and, having found a novel situation, to learn how to control it. . . . In slightly different words, man is always aiming to achieve some goal and he is always looking for new goals. . . . My contention is that man enjoys performing these jointly innovative and cohesive operations. Together, they represent an essentially human and inherently pleasurable activity." As already discussed, with this reference to "new goals" Pask explicitly moved beyond the original cybernetic paradigm with its emphasis on mechanisms that seek to achieve predefined goals. This paragraph also continues with a definition of "control" which, like Beer's, differs sharply from the authoritarian image often associated with cybernetics (76): " 'Control,' in this symbolic domain, is broadly equivalent to 'problem solving' but it may also be read as 'coming to terms with' or 'explaining' or 'relating to an existing body of experience.' " Needless to say, that Pask was in a position to relax these definitions went along with the fact that he was theorizing and exploring human adaptive behavior, not attempting to build a machine that could mimic it. Musicolour, for example, was a reactive environment; it did not itself formulate new goals for its own performances.

  Pask's opening argument was, then, that "man" is essentially adaptive, that adaptation is integral to our being, and to back this up a footnote (76n1) cites the work of "Bartlett . . . Desmond Morris . . . Berlyn . . . Bruner . . . social psychologists, such as Argyll," and, making a connection back to chapter 5, "the psychiatrists. Here, the point is most plainly stated by Bateson, and by Laing, Phillipson and Lee [1966]." Of course, Bateson and Laing and his colleagues were principally concerned with the pathologies of adaptation, while throughout his career Pask was concerned with the pleasures that go with it, but it is interesting to see that he placed himself in the same space as the psychiatrists.12

  Pask's essay then focused on a discussion of "aesthetically potent environments, that is, . . . environments designed to encourage or foster the type of interaction which is (by hypothesis) pleasurable" (Pask 1971, 76):

  It is clear that an aesthetically potent environment should have the following attributes:

  a It must offer sufficient variety to provide the potentially controllable variety [in Ashby's terms] required by a man (however, it must not swamp him with variety—if it did, the environment would be merely unintelligible).

  b It must contain forms that a man can learn to interpret at various levels of abstraction.

  c It must provide cues or tacitly stated instructions to guide the learning process.

  d It may, in addition, respond to a man, engage him in conversation and adapt its characteristics to the prevailing mode of discourse.

  Attribute d was the one that most interested Pask, and we can notice that it introduces a metaphor of "conversation." An interest in conversation, understood very generally as any form of reciprocally productive and openended exchange between two or more parties (which might be humans or machines or humans and machines) was, in fact, the defining topic of all of Pask's work.

  Having introduced these general aesthetic considerations, the to the essay then devoted itself to descriptions of the two machines—Musicolour and the Colloquy of Mobiles—that Pask had built that "go someway towards explicitly satisfying the requirements of d."We have already discussed Musicolour, and we can look at the Colloquy later. Here I want to follow an important detour in Pask's exposition. Pask remarks that (77) "any competent work of art is an aesthetically potent environment. . . . Condition d is satisfied implicitly and often in a complex fashion that depends upon the sensory modality used by the work. Thus, a painting does not move. But our interaction with it is dynamic for we scan it with our eyes, we attend to it selectively and our perceptual processes build up images of parts of it. . . . Of course, a painting does not respond to us either. But our internal representation of the picture, our active perception of it, does respond and does engage in an internal 'conversation' with the part of our mind responsible for immediate awareness." This observation takes us back to the theme of ontology in action: what difference does ontology make? It seems that Pask has gone through this cybernetic analysis of aesthetics only to conclude that it makes no difference at all. Any "competent" art object, like a conventional painting, can satisfy his cybernetic criterion d.So why bother? Fortunately, Pask found what I take to be the right answer. It is not the case that cybernetics requiresus to do art in a different way. The analysis is not a condemnation of studio painting or whatever. But cybernetics does suggest a new strategy, a novel way of going on, in the creation of art o
bjects. We could try to construct objects which foreground Pask's requirement d,which explicitly "engage a man in conversation," which "externalize this discourse" as Pask also put it—rather than effacing or concealing the engagement, as conventional art objects do. Cybernetics thus invites (rather than requires) a certain stance or strategy in the world of the arts that conventional aesthetics does not, and it is, of course, precisely this stance, as taken up across all sorts of forms of life, that interests me.